Zalman 9500: "Better" than water?
 

Zalman 9500 LED

Hey guys first up, will say this this review is done on my local website. Aint doing advertisement here, more of looking for constructive criticism about the reviewing method.

The guy who's done the review is "under my wing" in a sense, been whipping him into shape ala "jaydee" style till he produced this review which was I felt was well enough to be posted online and risk being posted here especially.

First off, I know that the thermal sensor is bullshit using an Aerogate. that makes the temps all questionable especially here on ProCooling.

Secondly, no repeated mountings to average out the data individually on the three heatsinks.

What else would there be?

I am still in the process of fixing up the die sim that will be used to test heatsinks ( I dare not test waterblocks) and its having the die itself being shaved down a couple of mm^2 before its ready.

Till then, please by all means have a go at the review and let me know what can be improved besides the two I've mentioned above so that heatsinks can still be tested before the die is ready.

As for the Zalman 9500. Its pretty decent, but not all that impressive. Prefered my older Hyper6 for sheer overclocking stability, but this guy does almost the same quieter.

Not bad for a first review. Certainly better than a "mainstream" review. A few problems:
1. No C/W chart. This makes it difficult to extrapolate results for different CPUs
2. Bad temp chart. This is difficult to read with accuracy.
3. Aerogate is has not been validated as accurate. Lay down $50-$100 for a decent set up if you plan on doing these reviews regularly.
4. The IHS has an impact on a coolers performance. This makes the comparison of HSFs slightly inaccurate, though I wouldn't dock you for it simiply because I don't like the concept of an IHS (anit crush pads on the side would do just fine).

This is all IMHO of course. I hope it helps.

Noise level
Needs more material in this area. Due to subjective nature of this area he needs data. Spend more time in this area and compare to other products. Have a comparision graph. Obviously this product is targeted towards the silent enthusiast and this would help that audience.

Test result and organization
This is the most important part of the review. Everyone talks about the results and will forgive a okay review if it has good test results section.
That graph is horrible to read and compare temps. Needs better increments or something to help out. Maybe have larger graphs and bars?

Multiple graphs instead of a single one. He should have one graph for CNP 9500 at different settings and compare the different hotizontal and vertical setups. Use the better setup to compare to other products.

Mhz/voltage chart is a nice touch. I haven't seen that yet. He should setup his Delta T results and sound levels like this or on a table. Just minor organization helps

Testbed/setup

Procooling loves to eat this up

As maxsaleen pointed out, the IHS can change results so he should've just done a test with IHS before removing it. Most AMD 64 use the IHS so he should have IHS test data.

Repeat mountings. Helps eliminate bad results due to a bad mount. Usually to keep from getting bad results but the results look normal. I have no beef here because most mainstream reviews don't do this.

Aerogate, not the best but good enough for mainstream review. This is not really indepth testbed and using aerogate for delta T is basic enough. No beef here....

He should test the vertical setup by moving his card to a lower slot. By having the intake directly above the VGA card, it is getting heated air from the back of the card and also blocking intake. That affects temps a lot. A sneak peek into how that affect gpu temps would be a nice little extra surprise.

Overall very good review but lacks info on noise levels and lack of organization of test data and results hurts the review. Hurts the review a lot.

Not entirely sure taking the IHS off was a good idea. Many heat sinks and water blocks mounting hardware is designed with the IHS on. Taking the IHS off lowers the heat sink and may reduce mounting pressure.

Looks like you guys know what is wrong with the test setup.

Were did the "better than water cooling" part come from? Did imiss something in the review?

It's an integral part of Zalman's marketing literature for this heatsink product...or it used to be up until actual release.

Oh that's right, I was thinking that was Thermaltake for some reason.

I'd say once you have a die sim, you'll be much closer to testing water blocks; all you need is to be able to measure the water flow rate and the pressure drop.

The die sim is really important, but it's not a complete solution to an accurate temperature measurement.

So much more to cover: join the club: wbta.us . Feel free to PM / email me.

bigben2k: Thanks for the invite to WBTA, I've been on here for ages so I've seen that attempt since its inception. i'll take a look and join in once I've got the die sim up and ready and got something to discuss about! :)

jaydee: yeah Zalman initially claimed that the 9500 was better than watercooling in their marketing blitz. Even Xbitlabs "claimed" in their testing that it beat the Gigabyte w/cing kit which is very dubious.

everyone else: Cheers for the suggestion regarding the noise factor. Will get a sound meter to accurately measure noise of the fans at the different rpms.

As for the IHS issue. Isnt there a flip side to this, removing the IHS we get more accurate temperatures closer up to the core. With the IHS on, half the time I've found the temperature gradient across the IHS to be relatively high with the edges showing quite a bit lower temps than the core.

With the smaller contact patch, we get back into the old AthlonXP vs. P4 debate on heatsinks dont we.

anyhow, I'll see how else we can get the standard up a bit more :)

Thanks fellas

I understand the better temp readings with removal of IHS but that test is based off a running setup. IT should be treated as an average setup. Removing the IHS removes a common variable from all AMD 64 cpus in an average setup except DTR's.

It doesn't matter here because we are comparing results using the same test bed.
But how will your results compare to the exact same test setup but with the IHS and gets different results? Which results will people morelikely to accept? The bread and butter IHS one. It's like testing P4 without the IHS. How many times you going to see that?

Note that others pointed out mounting presure designed with the IHS. This will affect performance on certain designs.

In the end that IHS is an important factor but i see how you want more accurate readings. Tough call.

BTW $145 for a heatsink?? Is that US dollars or some different currency.
I can get a good decent WCing setup for that.........

Singapore Dollars mate, its just too bloody steep. (SGD$145 = ~USD$86)

I understand what you're saying about comparative purposes. What if this setup was to stay the same. No changes made, the non-IHS CPU used, withe the same board, same everything. Only difference would be the heatsink and the ambient temperatures, and of course TIM application (TIM material would stay the same).

Wouldnt that then be sufficient in comparing several different heatsinks with past and present data ?

my only recommendation is:

DITCH those stupid rankings at the end. 9/10 EDITORS CHOICE ! Bs like every site has. I do feel reviews should be quite factual and let the facts speak for themselves. (PC had their own ranking like that a while ago, but sorta ditched it when it became clear that one writers oppinion is not = to what the facts present).

otherwise not too bad for a first try.

Actually, the WBTA is here to help from the very beginning! :cool:

Now that is something new to consider and honestly, something I do agree with Joe. I shall attempt to implement that immediately with the next few reviews.

bigben2k: *grin*, no worries mate I'll pop by once I have questions. Now its a matter of finding time to shave down those mm^2 and still do the reviews I have piling up on my table and carry on my normal work life and family life. So little time, so much to do!

Completely agree with Joe. Scrap the ranking. As for the IHS it really depends on the mounting system. If it was designed to have a set pressure rate with the IHS then the results are going to be flawed if the IHS is removed. Especially if the mounting design is height sensitive. The heat sink might dominate all others with no IHS but it might be worst than all others with the IHS or just the opposite.

Although another issue I have thought about in the past is wearing out the IHS's TIM layer and gradually having an effect on temps. I can't imagine the IHS was designed for continuous and repetitive mounts.

That is something that is plaguing my thoughts these days.

I've got a nice lil venice that when i first had it could do 2.85GHz @ 1.55v with a Hyper6. Now I'm having issues just getting it stable at 2.8Ghz with a Zalman CNPS7700Cu.

The effect of different heatsinks on overclocking is very well known but I've mounted & remounted a whole slew of waterblocks, evap heads & heatsinks on this lil venice quite a bit (close to 50~60 times) and the "feel" (I have no conclusively data here) is that the system's overall o/c has degraded over time and I'm beginning to suspect its the TIM joint between IHS & core.

I also understand the issue of the IHS. How about using the IHS, drilling a small hole in the side so that the probe wire can be trailed outof the there and then mounting the heatsinks as per normal. We'd still be getting the temps of the core (more accurate than the side of the IHS) and we wouldnt have to worry about the different heights and mounting pressure concerns.

suggest grooving the IHS for a TC per the Intel P4 specs and bonding one (40 ga) in,
then the TC + CPU can be caled as any other sensor
nothing like doing something the 'same' as everyone else (in industry);
"W" will still be an unknown however

use a single reference to establish an initial baseline, if that baseline is 'lost' the bench must be re-caled as something has changed
this baseline verification is a part of every test - but seldom acknowledged (or performed ?)

Not bad, not bad at all. I'll give it 4 flaming chihuahua's out of 5. ;) It does include a few of the silly "standard" review site bits that mean nothing but almost everyone includes for some reason, like worrying about how "reflective" the base is, but otherwise pretty well written.

Most of what would have been my comments have already appeared above. While trying to avoid flagging dead horses, I'll add just a couple of thoughts:

Yes, lose the rating scale. Not only is it purely subjective, as others have said, but it also paints you into a corner. You gave the 9500 4/5, so if the next heatsink you test performs twice as well you'll almost have to give it a 5/5, right? But then what happens if the next heatsink after that performs four times as well? Give it a 6/5? Or go back and change the 9500's score to a 3/5? It could get messy. If you want to give readers a simple way understand where you think this product ranks in comparison to others one alternative is to provide a separate page where you rank the tested products by whatever standard you choose. That way when something new comes out you can simply reshuffle the rankings without having to go back and edit the older reviews. (shameless self-promotion alert: SPCR's Recommended PSU Listing)


On the IHS issue I would vote for always leaving it on. Test the product in the way the vast majority of your readers will be using it, even if it does affect accuracy slightly. As long as the testing is comparable between HSF's, pure "accuracy" is actually less important.


Your results graphs definitely lead something to be desired, as others have said. Just hard to read overall. If you're not going to break it out as °C/W, then dT should be the most prominant result, so don't let it be lost in the mess of other numbers. I wouldn't hold it up as the model of perfection for heatsink testing, but here's our review of the 9500LED if you wanted to see a different way of presenting similar results.


Acoustic testing. Don't even bother trying to measure the noise. You won't be able to get equipment that can even come close for under $1000, and even then you'll likely find that your testing environment will need serious modification to produce meaningful results. You're actually better off just describing it subjectively, although your definition of "slight whining sound" is probably up for debate. At full speed our testing of the 9500 put it at 37dBa...too loud for us to even consider using at that speed.

We're not in spcr anymore, peoples' ear's aren't as sensitive. Looking at his setup it seems as if the heatsink is making very little difference in the noise level.

I'm also for leaving the IHS on since most people will run their setups with the IHS. As long as you use the same IHS it shouldn't have that much of an effect should it?

Well I don't agree. Most Wcing users will agree that the benefit of WCing is less sound so some people like stuff very silent. What is the focus of PA160 and 120s? Better performance at lower fan speeds for less noise.

Telling me it's 37 dBAs tells me a lot more than it has a soft whining noise.
I realize it's quieter than most heatsinks but not silent like how the reviewer made it out to be until I see 37dBAs. Kinda gives the impression the fan is a tad noisy but tolerable. Still needs other dBAs to compare to because dBA can vary test to test.

I like the Wcing pump review in french where he had all the soundfiles to compare the noise from each pump. That should be the standard. How come SPCR can spend $1000 on that but not capture more sound clips?

Dryfire, my point wasn't that they shouldn't try to talk about noise. It was more that the cost for measuring noise is prohibitive. To paraphrase the Procooling Testing Mantra: "If you're not going to do it right, don't do it at all." :dome: Subjective noise descriptions are essentially free, and can be really useful if you can compare the noise to sources that the readers likely have. For many people saying that a fan is, "louder than a AMD stock cooler" is actually more useful that saying that it's 27.5dBa@1m. It doesn't require anything special to describe the noise, well, maybe a little literary BS-ing ability will help, but ears are ears.

The trouble with trying to do acoustic measuring cheaply is that you end up getting useless numbers that then propagate is being hard numbers. You see it all the time on hardware sites: Someone buys a $30 Radio Shack SLM, discovers that they have to stick it 1" from the fan so it will read anything, and then they quote those results like they're gospel, without ever wondering why the readings disagree with what their ears are telling them. The WC'ing equivalent would be using an aquarium thermometer floating in your res as a method for comparing WB's. :)

For Sin22's site the payoff for the required investment just isn't there. I'd rather see them put the effort into a kick-ass die sim, and just use their existing free god-given acoustic measuring tools for noise.


ricecrispi, you need to come over to spcr-land more often: nearly every review for the past year has included downloadable sound files for comparison, in addition to the noise measurements.

What I meant is his total system noise is probably around 30 dba so it's not like he can really differenaite between the two very well. First thing he needs to do is build a very quiet system if he's going to do noise comparisons.

To me 37dba is a roar not a whine.

I forgot SPCR had those too. It's has been a while since I read one of those reviews. Just not focusing on that area any longer because you guys got it covered like a hammer on the head of nail.

I think that desribing noise is too subjective. What he called a whisper was actually quite loud. If he heard a nexus fan on a heatsink he probably couldn't hear anything. That's a true whisper.

Numbers help but they are just to compare. Soundfiles real solve issues of past sound noise levels testing because people can determine themselves by listening.

So what is reasonable price equipment that non SPCR sites use to test sound levels? I understand SPCR has this niche area covered but Sin22 review site can't be competitive with other review without a good sound noise level section. No one can compete with SPCR at that level.

A db number is completly useless without knowing the pitch.

Exactly. 2.5 khz at a low level is much more audible than the same level of, say, a 10k whine.

But if a measurement is given in db(A), the measurement itself is weighted for the threshold of hearing. So values given in dbA are usually a reasonable representation of perceived volume.

http://en.wikipedia.org/wiki/Equal-loudness_contour

http://en.wikipedia.org/wiki/DB%28A%29